ta
The AI Scientist-v2
Automated ML research via agentic tree search
AT A GLANCE
Intended use: Research experiments on open-ended automation of machine-learning discovery.
Pricing model: Free.
Public-source review: . Documentation checked; product performance was not tested.
LIMITATIONS
- The authors state that v2 does not necessarily outperform v1, particularly when v1 has a strong starting template.
- Executes generated code; use a controlled environment and budget for GPU and model usage.
- Uses the AI Scientist Source Code License, including an AI-use disclosure requirement for generated papers.
FEATURES
- ▸ Explores experiment alternatives through an agentic tree-search workflow.
- ▸ Connects ideation, experiment execution, result analysis, and manuscript drafting.
What it is
Sakana AI's research system uses agentic tree search to generate hypotheses, run machine-learning experiments, analyze results, and draft papers without human-authored experiment templates.